On the Pathwidth of Planar Graphs Omid Amini, Florian Huc, Stéphane Pérennes

Home , Bipartite graph, Degree (graph theory), Outerplanar graph, Pathwidth, Planar graph, Polyhedral graph

To cite this version:

Omid Amini, Florian Huc, Stéphane Pérennes. On the Pathwidth of Planar Graphs. [Research Report] 2006, pp.6. inria-00082035

HAL Id: inria-00082035 https://hal.inria.fr/inria-00082035 Submitted on 7 Jul 2006

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. INSTITUT On N A Omid TION the AL Amini P DE athwidth RECHERCHE — Florian Thème N° Juin EN ???? Huc 2006 of INFORMA COM — Planar Stéphane TIQ UE ET Pérennes Graphs EN A UT OMA TIQ UE

ISSN 0249-6399 ISRN INRIA/RR--????--FR+ENG

On the Pathwidth of Planar Graphs

Omid Amini ∗ † , Florian Huc ∗ ‡, Stéphane Pérennes ∗

Thème COM — Systèmes communicants Projets MASCOTTE

Rapport de recherche n° ???? — Juin 2006 — 6 pages

Abstract: Fomin and Thilikos in [5] conjectured that there is a constant c such that, for every 2- connected planar graph G, pw(G∗) ≤ 2pw(G)+c (the same question was asked simutaneously by Coudert, Huc and Sereni in [4]). By the results of Boedlander and Fomin [2] this holds for every outerplanar graph and actually is tight by Coudert, Huc and Sereni [4]. In [5], Fomin and Thilikos proved that there is a constant c such that the pathwidth of every 3-connected graph G satisﬁes: pw(G∗) ≤ 6pw(G) + c. In this paper we improve this result by showing that the dual a 3-connected planar graph has pathwidth at most 3 times the pathwidth of the primal plus two. We prove also that the question can be answered positively for 4-connected planar graphs. Key-words: planar graphs, pathwidth

∗ Projet Mascotte (cnrs, inria, unsa), INRIA Sophia-Antipolis, 2004 route des Lucioles BP 93, 06902 Sophia-Antipolis Cedex, France † École Polytechnique, 91120 Palaiseau ‡ This author is partialy supported by Région Provence Alpes Côte d’Azur

Unité de recherche INRIA Sophia Antipolis 2004, route des Lucioles, BP 93, 06902 Sophia Antipolis Cedex (France) Téléphone : +33 4 92 38 77 77 — Télécopie : +33 4 92 38 77 65 A Propos de la Pathwidth des Graphes Planaires

Résumé : Fomin et Thilikos[5], après avoir démontré que la pathwidth de tout graphes planaires 3-connexe est au plus 6 fois celle de son dual à une constante près, ont conjecturé que pour tout graphe planaire biconnexe G, pw(G∗) ≤ 2pw(G) + cte. D’après Boedlander et Fomin [2] cela est vrai pour tout graphe outerplanaire. De plus cela est exact d’après Coudert, Huc and Sereni [4]. Dans cet article nous améliorons le résultat de Fomin et Thilikos en montrant que la pathwidth de tout graphe planaire 3-connexe est au plus 3 fois celle de son dual plus 2. Nous démontrons également que la conjecture est vrai pour tout graphe planaire dont le dual est 4-connexe. Mots-clés : graphes planaires, pathwidth On the Pathwidth of Planar Graphs 3

1 Introduction

A planar graph is a graph that can be embedded in the plane without crossing edges. It is said to be outerplanar if it can be embedded in the plane without crossing edges and such that all its vertices are incident to the unbounded face. For any graph G, we denote by V (G) its vertex set and by E(G) its edge set. The dual of the planar graph G, denoted by G∗, is the graph obtained by putting one vertex for each face, and joining two vertices if and only if the corresponding faces are adjacent. Note that the dual of a planar graph can also be computed in linear time. The notion of pathwidth was introduced by Robertson and Seymour [9]. A path decomposition of a graph G = (V, E) is a set system (X1, . . . , Xr) of V such that r 1. Si=1 Xi = V ;

2. ∀xy ∈ E, ∃i ∈ {1, . . . , r} : {x, y} ⊂ Xi;

3. for all 1 ≤ i0 < i1 < i2 ≤ r, Xi0 ∩ Xi2 ⊆ Xi1 .

The width of the path decomposition (X1, . . . , Xr) is max1≤i≤r |Xi| − 1. The pathwidth of G, denoted by pw(G), is the minimum width over its path decompositions. Computing the pathwidth of graphs is an active research area, in which a lot of work has been done (Fer a survay see for instance [8]). Govindan et al. [6] gave an O(n log(n)) time algorithm for approximating the pathwidth of outerplanar graphs with a multiplicative factor of 3. For biconnected outerplanar graphs, Bodlaender and Fomin [2] improved upon this result by giving a linear time algorithm which approximates the pathwidth of biconnected outerplanar graphs with a multiplicative factor 2. To do so, they exhibit a relationship between the pathwidth of an outerplanar graph and the pathwidth of its dual. More precisely, the following holds.

Theorem 1 (Bodlaender and Fomin [2]) Let G be a biconnected outerplanar graph without loops and multiple edges. Then pw(G∗) ≤ pw(G) ≤ 2pw(G∗) + 2.

Since the weak dual of an outerplanar graph (which can be computed in linear time) is a tree and there exist linear time algorithms to compute the pathwidth of a tree [11], this yields the desired approximation. Coudert, Huc and Sereni in [4] improved this result by proving the following theorem: Theorem 2 (Coudert, Huc and Sereni [4]) For every biconnected outerplanar graph G, we have pw(G∗) ≤ pw(G) ≤ 2 pw(G∗) − 1 and all the values in the interval [pw(G∗) , 2 pw(G∗) − 1] can be the pathwidth of G. Simultaneously Coudert, Huc and Sereni state the following question as an open problem in [4] and Fomin and Thilikos conjectured it in [5] :

1 ∗ Conjecture 1 ([5],[4]) Is there a constant c such that, for every 2-connected planar graph G, 2 pw(G )− c ≤ pw(G) ≤ 2pw(G∗) + c? It is worth noting that this conjecture is motivated by the following result about the treewidth, conjectured by Robertson and Seymour [10] and proved by Lapoire [7] using algebraic methods (notice that Bouchitté, Mazoit and Todinca [3] gave a shorter and combinatorial proof of this result).

Theorem 3 ([7]) For every planar graph G, tw(G) ≤ tw(G∗) + 1.

Fomin and Thilikos made an even stronger conjecture :

Conjecture 2 ([5]) There is a constant c such that for every 2-connected planar graphs G of treewidth ∗ m at least m, pw(G ) ≤ m−1 pw(G) + c

RR n° 0123456789 4 Omid Amini , Florian Huc , Stéphane Pérennes

This conjecture does not hold. Indeed we can slightly modify the examples given in [4]. They are example of biconected outerplanar graphs G such that pw(G) = 2pw(G∗) − 1. We modify this family of graph by plugging a 3 × 3 grid on a face. This can be done without changing the pathwidth, whereas the treewidth increases from 2 to 3, so the equation is no longer satisﬁed. The following theorem improves the previously known bound for 3-connected planar graphs.

Theorem 4 (Main theorem) For every 3-connected planar graph G we have pw(G) ≤ 3 pw(G∗) + 2

Actually our methods prove that the conjecture holds for every 4-connected planar graph.

2 Main Theorem

In this section we present the proof of our main theorem. We will use the following notations: Given a graph G = (V (G), E(G)) of maximum degree ∆(G), we will note V (G∗) either its face set or the vertex set of its dual and by fG, eG and nG respectively the number of faces, edges and vertices of G. Given a set A, by P(A) we denote the family of all subsets of A.

Deﬁnition 1 Let G and H be two graphs. A connected map from G to H is a map σ : V (G) → P(V (H)) from vertices of G to subsets of vertices of H satisfying the following two properties:

1. for every v ∈ V (G) , σ(v) is connected. 2. for every adjacent vertices v, w ∈ V (G), σ(v) ∪ σ(w) is also connected σ is of degree at most k if it also satisﬁes

• ∀w ∈ V (H) we have |σ−1(w) := {v ∈ V (G)|w ∈ σ(v)}| ≤ k

Lemma 1 Let G and H be two graphs. for any connected map σ of degree at most k from G to H, we have:

• pw(G) ≤ k pw(H) + k − 1 • tw(G) ≤ k tw(H) + k − 1

Proof • Given a path-decomposition of H of width `, applying σ−1 on bags of our decomposition gives one of width k.` + k − 1 for G. This can be easily verified using the properties of σ listed above. • Same proof gives the result for tree-width. From now on we suppose G to be a 3 vertex connected planar graph. We aim to find a low degree connected map from G∗ to G. A Face-To-Edge assignment is a system of distinct representatives for faces of G. In other words, a Face-To-Edge assignment is a function τ such that we associate to a given face F of G an edge τ(F ) = (v, w) ∈ E(F ) ⊂ E(G), in such a way that two different faces are associated to different edges. Given a Face-To-Edge assignment, the map σ : G∗ → P(V (G)) associates to every vertex F of G∗ (face of G) the subset V (F ) \ V (τ(F )).

Proposition 1 The so deﬁned map σ is connected.

INRIA On the Pathwidth of Planar Graphs 5

Proof Two faces F1, F2 sharing an edge e can’t be both associated to e since they are associated to diﬀerent edges. Consequently σ(F1) ∪ σ(F2) is connected. Given a Face-To-Edge assignment, let H be the subgraph of G consisting of non selected edges; i.e. ∗ H = G \ {τ(F )|F ∈ V (G )}. Using Euler’s Formula (fS + nS = eS + 2) we know that H contains exactly n − 2 edges. We have

−1 Proposition 2 ∀v ∈ V (G), |σ (v)| = degH (v)

Proof A selected edge (an edge of G \ H) should be associated to one of the two faces containing it. Given a vertex v of G of degree d, it appears exactly in d faces. Suppose r edges incident to v are selected, so they should be associated to exactly r faces incident with v. v doesn’t appear in |σ(v)| of these faces, −1 and appears in |σ(v)| of other faces incident to v. So |σ (v)| = d − r = degH (v) Corollary 1 σ is of degree at most ∆(H). Remark that the average degree in H is always leq2. Deﬁnition 2 We call H ⊂ G a nice subgraph if it has n − 2 edges and such that we can ﬁnd a Face-To- Edge assignment τ with τ(F ) ∈ E(G) \ E(H).

Deﬁnition 3 Given a graph G, we call adjacency graph, the bipartite graph A on vertex set ( V (G∗) ∪ E(G \ H) ), with an edge between a vertex of V (G∗) (i.e. a face F of G ) and an edge of G \ H if this edge belongs to F .

Corollary 2 Let H be a nice subgraph of G of max-degree ∆. Then we have pw(G∗) ≤ ∆ pw(G) + ∆ − 1

We will need the following theorem of Barnette [1]: Theorem 5 (Barnette) Every 3-connected planar graph has a spanning tree of max-degree 3.

Corollary 3 Every 3-connected planar graph G contains a nice subgraph of max-degree 3

Proof By Barnette’s theorem there exists a sub-forest H of G of max-degree 3 containing n − 2 edges (i.e. is a spaning tree minus an edge). We want to prove that such a subgraph is nice by applying Hall’s matching theorem to the adjacency graph A between faces of G and edges of G \ H. Given a set of faces {F1, . . . , Fi} we should prove that in A the corresponding set has at least i neighbors. Considering the planar graph S obtained by taking the union of Fi, we have:

∗ • fS ≥ i + 1 (because G is connected)

• fS + nS = eS + 2 (Euler’s formula)

We conclude eS − (nS − 1) ≥ i. As H is a forest the number of edges of H incident with some vertex of this subgraph is at most nS − 1. So the hypothesis of Hall’s theorem is always satisﬁed. This proves that H is a nice graph. As a corollary we have Corollary 4 There exists a connected map σ : G∗ → G of degree at most 3. As a results from corrolary 2 and 4 we have our main theorem:

Theorem 6 For every 3-connected planar graph G we have pw(G) ≤ 3 pw(G∗) + 2

Furthermore our method proves the conjecture for planar graphs whose dual has an Hamiltonian path :

RR n° 0123456789 6 Omid Amini , Florian Huc , Stéphane Pérennes

Theorem 7 If G has a Hamiltonian path, we have pw(G) ≤ 2 pw(G∗) + 1 Proof The Hamiltonian path gives a nice subgraph of G of max-degree 2. Corollary 5 If G∗ is 4-connected then we have pw(G) ≤ 2 pw(G∗) + 1 Proof Thomassen proved in [?] that every 4-connected planar graph has a Hamiltonian cycle. Then by last theorem we have the result. Acknowledgments. The authors would like to thank David Coudert, Jean-Sébastien Sereni and Stéphan Thomassé for interesting discussions.

References

[1] D. Barnette. Trees in polyhedral graphs. Canad. J. Math., (18):731–736, 1966. [2] H. L. Bodlaender and F. V. Fomin. Approximation of pathwidth of outerplanar graphs. J. Algorithms, 43(2):190–200, 2002. [3] V. Bouchitté, F. Mazoit, and I. Todinca. Chordal embeddings of planar graphs. Discrete Math., 273(1-3):85–102, 2003. EuroComb’01 (Barcelona). [4] D. Coudert, F. Huc, and J. S. Sereni. Pathwidth of outerplanar graphs. Research Report INRIA, (5804), 2006. [5] F. V. Fomin and D. M. Thilikos. On self-dulaity for pathwidth in polyhedral graph embeddings. Reports in Informatics, (316), 2006. [6] R. Govindan, M. A. Langston, and X. Yan. Approximating the pathwidth of outerplanar graphs. Inform. Process. Lett., 68(1):17–23, 1998. [7] D. Lapoire. Structuration des graphes planaires. PhD thesis, Université de Bordeaux, France, 1999. [8] B. Reed. Treewidth and tangles: an new connectivity measure and some applications. In R. A. Bayley, editor, Surveys in Combinatorics, pages 87–162. Cambridge University Press, 1997. [9] N. Robertson and P. D. Seymour. Graph minors. I. Excluding a forest. J. Combin. Theory Ser. B, 35(1):39–61, 1983. [10] N. Robertson and P. D. Seymour. Graph minors. III. Planar tree-width. J. Combin. Theory Ser. B, 36(1):49–64, 1984. [11] P. Scheﬄer. A linear algorithm for the pathwidth of trees. In R. Henn R. Bodendiek, editor, Topics in Combinatorics and Graph Theory, pages 613–620. Physica-Verlag Heidelberg, 1990.

INRIA Unité de recherche INRIA Sophia Antipolis 2004, route des Lucioles - BP 93 - 06902 Sophia Antipolis Cedex (France) Unité de recherche INRIA Futurs : Parc Club Orsay Université - ZAC des Vignes 4, rue Jacques Monod - 91893 ORSAY Cedex (France) Unité de recherche INRIA Lorraine : LORIA, Technopôle de Nancy-Brabois - Campus scientiﬁque 615, rue du Jardin Botanique - BP 101 - 54602 Villers-lès-Nancy Cedex (France) Unité de recherche INRIA Rennes : IRISA, Campus universitaire de Beaulieu - 35042 Rennes Cedex (France) Unité de recherche INRIA Rhône-Alpes : 655, avenue de l’Europe - 38334 Montbonnot Saint-Ismier (France) Unité de recherche INRIA Rocquencourt : Domaine de Voluceau - Rocquencourt - BP 105 - 78153 Le Chesnay Cedex (France)

Éditeur INRIA - Domaine de Voluceau - Rocquencourt, BP 105 - 78153 Le Chesnay Cedex (France) http://www.inria.fr ISSN 0249-6399